I'm using a prebuilt Q6600 computer with a locked down "can't be overclocked" Intel G33 motherboard. Presently it is stable as the Rock Of Gibraltar with the stock voltage settings when running at 3.2 GHz. This was achieved by taping ECU pin BSEL2 to change FSB from 1066 to 1333 as shown here:



And then by using "SetFSB" by Abo to boost me from 3.0 to 3.2 Ghz by utilizing Clock Generator CV183APAG. I determined this selection through trial and error back in December. I could use "SetFSB" on the 1066 bus to get me to 3.0 stable and a bit more, maybe 3.1+ but that was not stable. So the FSB change to 1333 is netting me some more overclock net.

I'd like to bring the ECU to 3.4 - 3.5 GHz. Beyond that it appears that my small case / air cooled concept will tend to get noisier, it appears. I'd like quiet. So my next goal is the 3.4 to 3.5 range. If I can keep that cool and quiet I will try for more. If it is too loud I'll wind it back a bit. If it's just right then I am done.

Those of you with motherboards running adjustable ECU voltage (Vcore) can just type in or arrow up a setting.

So I decided to see how the Motherboard itself "knows" what voltages to use, what FSB to use if you change ECU chips. Well it does this via the BIOS and with various settings on the motherboard and components.

On the ECU, a group of pins are either 1 or 0 (high or low) and the combination code is the voltage setting that Intel programmed in at the factory. My chip has a Vcore under load of 1.2625 Volts. Looking in the Intel documentation (#315592-005) on page 15 I see that 1.2625 volts comes from ECU pins VID1 through VID 6 being set to: 0,0,1,1,1,0 by Intel when produced.



So these ECU pins can be used to adjust Vcore directly, and with much greater precision than via the BIOS.

But they are hard to reach and if you want to iteratively tweak to find good settings, well this would be quite a PITA.

So where do these pins go? Well they connect to the motherboard and it's components. One is the programmable Synchronous Buck Controller which on LGA775 boards is generally near the top edge and above and to the right of the ECU socket and above and to the left of the first memory slot.

Here is where mine is located:



A close up:



Well this component controls the MOSFETs that serve Vcore to the ECU. It's the voltage boss. Well it has the VID inputs so that it know what voltage the ECU desires. This is much easier to fiddle with than the ECU pins.

Also, the chip has a leg that measures the Vcore. If you connect a suitable resistor between that leg (red) and ground you can directly adjust Vcore with an adjustable pot. There are a few hookup options to try out.



At the moment I am still pondering. I am trying to see what voltages seemed to work for others. I am noticing HUGE variations so i will likely just guess and then see how far it overclocks iteratively. For example if I connect VID3 with VID2 (brings the pattern from 0,0,1,1,1,0 to 0,0,0,1,1,0) which will tell the chip to set a Vcore of 1.3125 volts. Or connect VID 4 & VID 3 to VID2 setting all to 0 for 1.4125 volts. The increments are 0.0125 volts from 0.8500 to 1.6000 volts and are charted in Intel's PDFs. I may need to ground those pins to change a 1 to a 0, I have not tried this yet, it is doable in some manner. The idea is to be able to change those pins to 0 or 1 (low or high) as desired.

You can connect the pins a few different ways (if that turns out to be the best way). For example, by using a conductive ink pen at a suitable point. Such ink or paint is easily removable.

And of course if you solder to the FB signal you can load in a pot. This chip also sets Vdroop...which some hate and some love. It also can change the difference between idle and load voltages and can affect boot characteristics.

Wow -

I'm impressed with your progress on this to date.

No comments but to let you know I'm reading this out of true interest. This rig's the only one I have that OCs, but it's by far not the only one I own...

Thanks Jaafaman. As of tonight I am chugging away at a Stable 3.4 GHz On My 2.4 Ghz "Non-Overclockable" Prebuilt! I was at 3.2 Ghz up until now.

I had a few minutes after the Superbowl to work on my VDroop. This is the voltage drop you find under high load from the high load ECU "vcore" voltage. This is not a bad thing as it avoids other power supply issues and helps cut costs by allowing simpler, cheaper components to be used. Such as cheaper capacitors.

I thought I would reduce this droop a bit before I would try raising my VCore up from 1.2625 volts to some higher figure. First let me use most of my present Vcore was my thought. I don't want to eliminate Vdroop as that has side effects. By doing this first I could get some more overclocking without driving up heat too much as the actual peak voltage was not exceeded.

In my case:

Low Load / High Load Spec / High Load Actual / ***High Load Actual After Mod***

1.16 Volts / 1.2625 Volts / 1.19 - 1.23 Volts /***1.24-1.25***

So I am still seeing a drooping of about 0.01 - 0.02 volts rather than about 3 times that amount. This can also help startup up of highly overclocked rigs in some case, I understand. Attention D!!! I used Everest Ultimate to track my ECU voltage.

I did this mod in a simple manner. I added an electrical connection between two of the Synchonous Buck Controller's pins, with some resistance. These two pins are the CSSUM and CSCOMP on the little square device in the picture which is the SBC.

This mod was done by using a pencil and took about 60 seconds per try. With the computer off and power plug out I penciled the component shown a bit, so that the two soldered points had some connection to one another - a shunt. Pencil lead has graphite and clay in it and can conduct but has some resistance. Booted up and checked voltage...did it some more. Third time is the charm. I decided to stop with about a 0.01-0.02 volt Vdroop so that I could have a gain yet still have some protection. Don't get greedy! This is also easy to reverse as you just clean off the component with the computer off of course. If you go too far you would have no more vdroop at all which is not what the power supply was designed for...if this happens just clean the part and try again. I will also have to see how permanent this mod is, I would not want it to wear off in a few months.



Certainly I need to do more stability tests. But so far it seems that the excessive vdroop was holding me back a fair bit in my overclocking. If I am at 3.4 Ghz with such low voltage and able to process photos with all four cores very active along with certain benchmarks that is a good sign. I may have to revise my "Stop At 3.4 - 3.5 Plan" a bit as I am getting there too easily. I certainly won't need much extra Vcore to hit 3.5 it seems. Ultimately heat will be the thing that cuts me off I suppose.

I guess that is what happens when you overclock...you always want more!

Ja, natürlich.

Had to back my E6600 from 3.5 to 3.45GHz to accomodate the installed cooler. The cooler that's coming with the Quad is a heatpiped ZEROTherm rated for 130W, so it ought to handle a Q6600 in its sleep.

I tried the VDroop Mod just a week ago with near-perfect results. Used a solid graphite #6B pencil, and in one shot pegged it dead on. What a stroke of luck - the voltage stays within 0.0005 of what I set it for - at some points to the high side, at others to the low, but with such an insignificant difference I can just say:
"It goes to and stays where I set it"
Figure this ought to make taming the Quad quite a bit easier. (Still wishing I'd paid for Express Delivery - damn I'm getting antsy!)

Good job on tracking down the corresponding resistor on the G33.

Keep it up - you're doing quite well...
(edit: your approach is so "old school" and reminds me of so much I couldn't drop this thread now if I had to...)

Well what I try to do is ignore most forum hype and go back to basic principles and try things out. This works incredibly well if you modify cars!! Since it is effective and you cream all the folks who do the popular forum mods without understanding the underlying reasons for a modification. Or even if they work at all.

So for example by googling the device number from the photo, I had the manufacturer's PDF of all of it's characteristics a moment later. Identifying where to place the shunt took only another few minutes. It goes in the same place on any motherboard using a SBC concept that is even close to similar. Once you know the pins you just find a candidate component. So I found the right place very, very easily. It's safe to try out since your PC will run fine if you overdo it, but may have increased instability if you benchmark it such that all cores are super loaded down hence drawing lots of power. Only way to know is to try it out. Seeing as how you are running tighter than me on Vdroop, I may try to add some more pencil. Since it is so darn easy. I just used whatever was in my mechanical pencil in my desk drawer.

From what I am seeing/reading heatwise on the newer ECU chips, you can now go air cooled to much higher speeds than in the past. Much less need to go to water or whatever. I mean loaded with my photo work (which is often four cores at 100%) I run about high 40s to low 50s on ECU temps. And the fan is not all the way up. I'd prefer air cooled over water as it is so much cheaper and easier and allows easier access to PC components. I don't need to set a new benchmark record. The way ECUs behave is there is a certain degree of fairly easy overclocking...and then things get MUCH harder and lots of voltage and heat mods and magic is needed to go further and further. I want to stay on the easy path but go nearer to the point where things go uphill.

With the newer super coolers such as the one you mentioned, there is even more headroom. If you can trust rumours, some Nehalem (4Q08) chips run at under one volt so heat will be even lower than in the present 45 NM ECUs. That may mean that at some point (and maybe for the penryn) that heat is not the main roadblock in the same way it has been.

Oooh, far from ignoring it, I take advantage of it when I'm bored.

SLI has got to be, among a couple others associated with gaming, one of the most misunderstood technologies around. Now you'd think that a place like the SLI Zone would be where you'd go to get the correct "picture", but I swear that 98% of the people that helped create such a place must've long ago vaporized or something, cause there sure are some mixed-up folks passing advice these days.

So I show up at times with the proverbial "sharp stick" just to poke holes in their "theories", although I may stop this habit before long. It kinda loses its appeal by the time they all realize you actually know what you're talking about.

Kids today - ya gotta love 'em 'cause there's too many. You'll never really "win" with those odds...

My thoughts so far...

While I'm waiting for the Q6600 (no, not today either!), I tried your BSEL2 mod on the E6600 with one small change - used red tape.

Hmm...noticed on the E6600 specs that the miidle combo that's reserved on the Q6600 gives an FSB of 200 - looks like the die covers everything from 1.8GHz to 3.0GHz. Guess which I chose?

Interestingly enough, specific combinations of voltages and frequencies still result in the same operational temps, but...

I can run each frequency as much as 0.04V lower in some cases, with an average of about 0.025V. Stable at lower voltage = cooler running anyway.

Interesting phenomenon...

i still cant get my q6600 to 3ghz stable tried for way too long.

Well then something else besides the Q6600 ECU may be holding you back? It seems to respond quite well to OCing.

BTW I've done more torture testing today and the 3.4 GHz is holding nice and strong. It craps out at around 3.41+ or so with present mods. So backing it off about .1 GHz seems to be an adequate safety factor so far.

With how well the rig responded to just a sprinkle more net Vcore voltage, it seems like I won't need to add much more voltage to hit my goals.

Prior overclocking to 3.0+ starting with a 1066 bus showed me that the motherboard could take a good chunk more FSB speed, and the memory I added is fast at low voltage plus it is only two sticks rather than four (which is supposed to be a bit easier for the motherboard's memory controller to handle with stability when you push it). For example my memory is now running at just over 900 Mhz yet before I had it at about 985+ and it was still fine. Still the only way to know is to go ahead and try it out. I'll hit 985 on memory again at about 3.7 Ghz. So with the present adjustments open to me that may be one of the ceilings I run into if the air cooling keeps temps down with low fan noise at that speed. I'd have to do more motherboard volt mods to go much further...like on that chart (above) it becomes a project.

Here is the Intel Quad Vcore versus Pin Status chart. It's from Intel's Datasheet PDF on the Quads, page 15:

ftp://download.intel.com/design/processor...ts/31559205.pdf

Okay for those who'd like to add ECU voltage (vcore) to allow more overclocking and would like to do this at the ECU, here are some ideas. Just tape the pads as shown. Use an exacto blade to make tiny pieces of tape after putting some vinyl electrical tape on a piece of glass or a mirror. Then load them into place. Many other combinations are possible, roughly every 0.01 volts, just see the chart above.

The Q6600 starts out with VID6 => VID0 at 0 1 1 1 0 0 so for example looking at the chart above to get to 1.3125 volts (+0.05) we need 0 1 1 0 0 0. This means that VID3 needs to change from a high value (1) to a low value (0).

I hope to be able to test out this stuff in a few days.

Woohoo...IT WORKED!!!

But not at first. I made a mistake and did not have the VID signals figures out correctly. Rather than just taping the selected VID pins, you need to connect them to ground which in that region is called Vss. There are plenty of nearby Vss pins so no worries there. So I had it backwards however there is a reference point issue that makes that tricky...in any event it had to be one way or the other so you just give it a shot. This is the process folks.

I am using rear window defogger repair paint which is a lacquer with copper dust in it. It comes in a tiny bottle and dries extremely fast. Now I have a 20 year supply since you use very little per application. I used a small paint brush most of whose bristles I trimmed away so that a very thin line can be drawn.

So now I am up 0.05 volts as I decided to try that first and maybe that will be enough for me. Since I am at the part of the overclock curve where you work harder and harder to get less and less result.

Old Vcore: 1.2625 Volts. New Measured Vcore: 1.3125 Volts. Combined with the previous Vdroop reduction...under load my ECU cores now see about 0.08 to 0.09 more volts under load.

My idle voltage changed from 1.16 to 1.21 Volts as expected.

Three points:

1) I need to see how much farther past my old 3.4 GHz I can now overclock. Which is the whole point of going through this hassle.

2) I can increase the gap between idle and load such that I can have a lower idle voltage if that is desired at some point. This is done at the SBC chip already discussed although this mod has not been shown. I can also easily achieve any useable voltage and with fine granularity.

3) I will redo the pictures for this way of pin modding. It's pretty easy.

Now....time to let her rip baby!

Hey I am an eyelash short of 3.6 GHz...but in marketing terms I AM there since they round up. I'm closer to 3.6 than Intel was to 2.4 which is where I started. That seems to be about the limit so far. I may repencil the Vdroop to get maybe a hair more, forget-about-it stability is the idea. Now I will use the thing for a day or two to see if it behaves and stays cool...temps seem fine so far.

So the additional 0.05 volts allowed my Q6600 nearly 200 MHz / .2 GHz more speed. As more voltage is added the increase in speed will become less and less.

Running the calculations shows that I would hit my prior memory speed limit at around 3.7 GHz on the stock 1.8 DDR2 voltage. So this may be a good place to stop so that everything has some safety factor. And so that further mods aren't needed (like modifying the motherboard for more memory chip voltage) and fan noise from air cooling stays nice and low.

Here are some unlabeled piccies of the Q6600 Vcore pin mod. In this case I elected to add 0.05 more voltage. I taped off the ECU leaving the pins to be connected exposed. I used the plastic card to tuck in the tape tight to avoid any leakage past the mask. I used a small paint brush...which in one of the shots you can see that I trimmed. This was to allow finer painting. It turns out that masking is optional as the trimmed brush could be controlled well enough without any masking. The mod works well and can help those without easier options to modify Vcore so that they can overclock higher...or if you want finer adjustments.

Hi, thank you so much for the picture. I was able to tape mod my q6600 G0 from 2.4 to 3.0 and used set fsb to move around 3.2, but i want to move a bit more say around 3.4, my mobo is dg33FB. Could u plz show me pic how to reduce Vdroop on this mobo?? i tried a lot but ur DG33 mobo pic is different from my DG33FB mobo, plz help. After doing this vdroop, i will move to changing core voltage.

Intel DG33FB mobo
Transcend 2GB 800 FSB DDR2 ram
q6600 2.4 oc to 3.2
400 W PSU, dual 12 v rail with 15 and 13 amp each.
Segate 360 GB HDD.
XFX 8800GT XXX edition clocked at 670.

thatz about my PC specs, and also do u think i need to replace my PSU, because my system lags in games and shutsdowing during intense gameplay.

Thanks